The Biden administration’s executive order on racial equity aims to address the entrenched disparities in US laws and institutions that have long denied people of color and other historically marginalized communities equal opportunities for success. To accomplish this objective, the executive order calls for federal agencies to identify the best methods for “assessing equity” in agency policies and actions. But doing so is no easy task, as centuries of structural racism have created a web of barriers to equal opportunity that requires multiple policies working together to dismantle.

Urban Institute experts offer many rigorous data tools and analysis strategies that can advance the executive order’s goals, and microsimulation models in particular could provide a powerful resource to disentangle structural racism’s barriers. But microsimulation models can produce reliable answers to questions about racial equity only if their underlying data and assumptions reflect the persistence of structural barriers confronting people of color.

What is microsimulation?

A microsimulation model is a computer program that mimics the operation of government programs and market processes for individual people, families, or businesses and projects how outcomes will change in response. Microsimulation models can explore challenging “what if” and “what would it take” questions, projecting how major reforms and the cumulative effects of incremental investments can narrow equity gaps.

Microsimulation models can also incorporate how different policies interact (like the minimum wage and benefit cliffs for safety net programs), how changing economic or demographic conditions cause variation in policy outcomes, and the long-term effects of policy interventions, even across generations. These models can highlight the diversity in experiences within and between groups by projecting the full distribution of outcomes, not just averages.

Existing microsimulation models can assess whether policies have equitable effects

The Urban Institute has built multiple microsimulation models that can help assess racial and ethnic inequities in federal programs.

• The Dynamic Simulation of Income Model (DYNASIM) projects the size of the US population for the next 75 years and estimates of its well-being, including employment, income, wealth, and health status. Recent results estimate that wage discrimination costs college-educated Black men nearly $1 million in lifetime income.
• The Social Genome Model (SGM) projects how the experiences and circumstances of children, youth, and young adults influence their well-being in adulthood. The model has shown that learning losses for kindergartners and ninth-graders associated with COVID-related school closures will likely diminish the future earnings of Latina girls more than those of white children.
• The Analysis of Taxes, Transfers, and Income Security (ATTIS) model examines the impacts of changes to safety net programs on individuals and families. Recently, it showed that COVID-relief stimulus policies considered over the summer could cut the poverty rate among Black people by one-third.
• Urban’s Health Insurance Policy Simulation Model (HIPSM) focuses on health insurance costs and coverage. It estimated that repealing the Affordable Care Act would increase the share of Black people without health insurance coverage to 20 percent as compared with 14 percent for white people.

Effective racial equity analysis must confront data challenges

These and other microsimulation models offer huge potential for equity assessments, but modelers must be mindful of how their data and assumptions capture racial and ethnic diversity and the structural or systemic conditions that can cause different experiences for people of color.

Before using any microsimulation model for equity assessments, careful consideration should be given to the quality and representativeness of the input data. Essential data should be disaggregated by race and ethnicity. If disaggregated data do not exist, a growing set of data science and statistical methods can help fill in gaps, but that must be done carefully and ethically.   

Urban’s models are built on well-respected, nationally representative datasets. In some cases, Urban researchers have adjusted and validated the data using supplemental survey and administrative datasets to ensure the datasets fully and accurately reflect the characteristics of the population. The SGM team recently rebuilt its foundational data file by merging more recent panel datasets together to ensure representation of Latinx individuals.

Even with disaggregated data, if models are not built to look at racial and ethnic inequities, they may produce misleading conclusions about policy effects. Some models may implicitly assume people of different races and ethnicities experience the effects of policy changes in the same way. Urban’s modeling teams strive to incorporate the effects of structural and systemic barriers into their models, comparing their approaches and findings to the evolving research literature, engaging with outside experts, and updating the models accordingly. As a result, when DYNASIM projects wealth gains and losses, it accounts for the effects of wage discrimination and for racial differences in the likelihood (and timing) of severe disabilities.

With mindful approaches to dataset construction, microsimulation models are powerful tools that can help answer the “what if” and “what would it take” questions about the potential equity impacts of existing and new policies. By using microsimulation models, federal agencies can begin to confront and disentangle the web of inequities in their programs and policies.

Gregory Acs, Jessica Banthin, Melissa M. Favreault, and Margery Austin Turner contributed to this post.